Video Recommendation Algorithm Based On Clustering And Hierarchical Model

With the national present study of 6G network,network entertainment represented by network video will achieve unprecedented development.While how to promptly and effectively recommend the user's interested network video to users is a crucial step in establishing the domestic market.With the advent of the intelligent era,users have overcome the barriers of the traditional network video and they can upload videos by themselves.This has led to an explosion in the number and types of the online video.In this case,the users will have to waste a lot of resources to find the online video that they are interested in.Therefore,users urgently need a high-experience video recommendation system to help them screen videos to alleviate the challenges brought about by the “information overload” problem of the online video.At present,the most popular recommendation system algorithm is a collaborative filtering algorithm.However,it still has problems such as sparse comment data,cold start,difficulty in expanding,and low user experience.Although some researchers have tried to improve it,the final result is still not ideal.The main reasons are as follows.Users' spontaneous uploads of network videos are irregular so it is difficult to build a model;The different ages of users have different interests and hobbies;The feature extraction of the network video is difficult;As users pay more attention to the personal privacy,they begin to reduce their evaluation of the videos;In order to release poor quality online videos,some unscrupulous online video providers illegally hire “water army” to criticize videos and so on.These problems lead to the current recommendation system is difficult to accurately dig out the user's preferences,which leads to the poor user experience and gradually loses interest in the use of the recommendation system.To solve the above problems,a video recommendation algorithm based on clustering and hierarchical model(VRBCH)was proposed to improve the performance of the recommendation system and further improve the user experience.Focusing on the user,to use the user's own attributes,user the matrix decomposition idea to fill in the missing items of the user attribute matrix,avoiding affecting the clustering effect of similar users due to privacy issues.similar users were obtained by analyzing the Affiliation Propagation(AP)cluster.The hidden characteristics of users' interests and the potential hobbies were found out through the attribute similar users.The system uses the video capture tool to collect web videos browsed by similar users,and looks for the path of the user to browse the video in the historical log,use the network video sorting to calculate the user's preference degree of a video,and then form a video pre-recommended set,avoiding the system directly randomly recommend to users from a large amount of video.In addition,the behavior of users can be used to mine potential interests that the user cannot express or express unclearly.Secondly,the user's historical data is used to calculate the user's preference degree of a video,and then the video's preference degree of a video is converted into the video's label weight,which avoids relying on the expert's group experience to predict the error of an individual's preferences.Finally,a recommendation list of videos was generated by using analytic hierarchy model to calculate the sort ranking of user preference with videos,which has a positive effect on improving the user experience.The experimental results on MovieLens Latest Datasets and YouTube video review text datasets show that the proposed algorithm has good performance in terms of Root Mean Square Error(RMSE)and the recommendation accuracy.